Modeling for Optimal PID Design

Even though PID controllers have been around for a long time, few industrial controllers use derivative action and the remaining PI controllers are often designed with formula-based tuning rules rather than through computer-based optimization. This paper will delve into some of the reasons behind these choices and show potential benefits of instead using software-based PID tuning. Three commonly used tuning rules are compared to software tuning with respect to performance and robustness over a large process batch. The study shows the importance of combining a fast, accurate modeling tool with the software design method and gives guidelines for future modeling tools with regards to desired process information. With moderate process knowledge it is possible to design controllers that are much closer to optimal than the three tuning rules, with significant performance improvements as a result.